To the authors’ knowledge, this is the first large case-control study to evaluate the relationship between statin use and RA risk in the US. Findings of this study demonstrate that statin use is associated with a modest increase in the risk of developing RA, but following adjustment for hyperlipidemia, this risk is diminished. Former statin users had an increased risk of RA versus current statin users. Patients who used statins for <1 year were found to have a slightly increased risk of RA versus those who used statins for ≥1 year. A slight dose-response relationship was observed for statin intensity and RA risk. However, the trends for statin use duration and intensity did not reach statistical significance, and no subgroup had a meaningfully increased risk following hyperlipidemia adjustment.
The reduction in risk seen with hyperlipidemia adjustment warrants further discussion. Prior literature suggests that hyperlipidemia may serve as an independent risk factor for RA [21], and the results of this study appear to support the notion that an increased RA risk may be due to hyperlipidemia and not statin use. However, it is important to consider that many patients receive a hyperlipidemia diagnosis during the initiation of statin treatment, and therefore, any attenuation in risk resulting from hyperlipidemia adjustment may not reflect a true association. With the exception of one prior study [19], no comparative study reported an analysis that specifically assessed the influence of hyperlipidemia/cholesterol values as an independent variable in adjustments. Therefore, it is unknown how hyperlipidemia impacted risk estimates in prior studies and if any consequential change in risk reflected a true association.
If the reduction in risk following hyperlipidemia adjustment is artifactual, the underlying mechanism for how statins may facilitate RA development is unknown. Some evidence suggests that statins may disrupt immune homeostasis by directly affecting T cells and inducing a shift from Th1 to Th2 immune responses [32, 33]. This may lead to B cell hyperactivity and trigger the production of autoantibodies [34, 35]. In addition, the burden of infections over time, compounded by a reduced Th1 response, may foster breakdown of self-tolerance, as infectious agents cannot be cleared as effectively as they would be under normal conditions [36]. Another possibility is that statins, being pro-apoptotic agents, may be capable of activating or exacerbating cellular apoptosis [37,38,39,40,41]. The release of endoplasmic and endonuclear antigens into circulation during cell death may induce the production of pathogenic autoantibodies [42, 43]. Cardiovascular risk factors, such as hyperlipidemia, may also play a role, as data has shown that patients who later develop RA tend to have significantly more atherogenic lipid profiles [44]. Therefore, statin use may be a proxy for hyperlipidemia, and thus, the unfavorable lipid conditions may be responsible for the association with increased RA risk, potentially through the chain of proinflammatory and immunogenic reactions [45,46,47]. These mechanisms for loss of tolerance would likely not induce autoimmunity on their own, but they may result in an earlier diagnosis or influence the progression of a condition, such as RA, in those who are already prone to the development of an autoimmune disease [18].
Of the studies that have previously investigated this topic, two found statins to have a harmful effect, three demonstrated a protective effect, and three observed no clear impact on the risk of incident RA. A retrospective cohort study of 511,620 statin users, which observed an increased risk of RA in statin users compared to non-users, also observed a greater RA risk for recent users compared to current users [19]; this aligns with the observation of an increased RA risk in former versus current users in the present study. With the exception of this one similarity, the present study does not seem to clearly support or refute the findings from any prior study that has investigated this topic, as the results have been largely heterogeneous overall. Therefore, it seems necessary to further discuss key methodological differences between the present and prior studies that could account for the conflicting data.
First, lack of non-statin reference groups: Two prior studies, showing a protective effect of statins against RA, compared statin users depending on persistence with treatment [20] and treatment intensity [22], precluding direct comparison with the current study.
Second, variability in definitions of outcome and exposure: Compared to prior studies, the present study used a more specific and exclusive RA definition that was selected in order to avoid misclassifying patients with conditions that present similarly to RA, as this may have been an issue in other studies. Sensitivity analyses with regard to the RA definition produced similar results. The definition of statin exposure also varied among studies. While the current study required at least two statin prescription fills for statin users, several prior studies required only one statin fill or did not define a minimum exposure requirement [19, 20, 22,23,24,25]. Some of these studies may have included users with near negligible periods of statin exposure, which may explain the inconsistencies in the findings. When the statin user requirement was changed to only one prescription fill, the results remained unchanged. However, in this study, statin users were identified by the number of statin fills, while other studies had to rely on prescription data without information on whether the statin was actually filled. Therefore, it is possible that a similar change to the statin user definition may have a meaningful impact on risk estimates in other studies. Lastly, there was variation in the considerations for potential lag-times. Assuming that statin exposure is unlikely to influence the risk of RA immediately after initiating treatment, RA cases diagnosed within 90 days of starting statins were excluded. Previous studies that did not account for this delay may have included patients whose RA symptoms predated the initiation of statin therapy. There is also an assumed delay between symptom onset and RA diagnosis. Chan et al. [48] estimated a median delay from RA onset to diagnosis of approximately 36 weeks; thus, many patients may have developed RA months or years prior to the recorded event date. However, several previous studies have considered the impacts of this delay, and the results remained largely unchanged in all studies [18, 19, 21, 22].
Third, consideration for potential confounders: Similar to four previous studies [18,19,20, 22], the current study excluded RA patients with DMARD use before the index date and non-RA patients with any DMARD use. Studies not excluding for DMARD use may have resulted in inaccurate classifications of RA and non-RA patients. This study considered a more comprehensive list of pre-existing conditions as potential confounders through the use of two comorbidity indices. Adjustments for the Charlson and Elixhauser Comorbidity Indices did not significantly alter the results. One other study accounted for a comorbidity index [24], but the reported comorbidity scores were much lower than in the present study. This was likely due to the younger age of patients in that study conducted within a military health care system and may explain the result discrepancies with the present study. When ten additional comorbidities were added to adjustments as independent variables, the risk remained unchanged. Consideration for the influence of hyperlipidemia, as previously mentioned, was another important factor that may explain the different results.
Simvastatin 10mg has been available over-the-counter since 2004 in the UK. Four studies were conducted in UK populations with study periods during this time [19, 22, 23, 25]. The current study was performed in the US, where statins are available via prescription only. Over-the-counter statin use may have led to an underestimation of statin exposure in both user and non-user populations. It is unknown how this may have impacted the findings in those studies.
Fourth, biases: Berkson’s bias may have influenced the results, as joint pains may have prompted a thorough work-up that resulted in statin initiation prior to a subsequent diagnosis of RA. Ascertainment bias may have also influenced the results. As noted by de Jong and colleagues [19], some patients who recently begin statin therapy may experience muscle-related symptoms [49], which may lead to a greater number of clinical visits, including rheumatology referrals. Thus, statin users may have been more likely to be diagnosed with RA than non-users.
Strengths of this study include its large and representative sample size with >16,000 RA patients, more than any prior study on this topic. The strong comparability due to exact matching of cases and controls likely reduced measured and unmeasured confounding. A large range of sensitivity analyses were performed, producing largely consistent results, which suggests minimal impact of residual confounders. Lastly, this study was specifically designed to assess RA as an outcome of statin use by comparing statin users to non-users. Many of the prior studies examined additional outcomes or did not utilize a non-user reference group.
There are several potential limitations of this study. First, this was a retrospective, observational study, and the use of statins was not randomized. In addition, statin use was determined via prescription fill data. However, as previously mentioned, many administrative data sources rely on prescription data without information on whether the prescription was actually filled, so others may have overestimated statin use even more than the current study. Case-control studies are limited in that they cannot be used to calculate incidence and are not suitable to imply causation. However, considering the low prevalence of RA among the general population, a case-control study seemed to have greater power to assess this association compared to alternatives, such as a retrospective cohort study. Limited information was available on rheumatoid factor or anti-cyclic citrullinated peptide antibodies and on lifestyle factors (e.g., diet, physical activity), which may be important risk factors for RA [50, 51]. Adjustments for smoking and obesity did not have an impact on risk estimates. However, smoking, obesity, and comorbidities were determined via diagnostic codes and could not be verified with clinical data, such as BMI, blood pressure, glucose levels, or inflammatory markers. Finally, duration and intensity assessments were limited by the large number of prevalent statin users in this study. The duration of statin use was unknown for some of these patients, so the actual number of individuals in both duration groups should be higher. Although the sensitivity analysis for intensity produced consistent results, it was not possible to accurately perform a cumulative exposure assessment to examine the impacts of differential statin intensity use over time.